LTN traffic prediction is too accurate

[bicyclize]

What is the source of the traffic count prediction in the LTN tool.
I downloaded the area from OSM and the baseline prediction is very accurate to real life counts. I attach a screenshot and a snippet from the local official model which is also verified by some manual counts.
Can you clarify, please how these numbers were calculated?

[dabreegster]

I'm quite surprised the numbers match. For UK maps, 2011 census data is used as the travel demand model -- that's super super old and only includes home-to-work trips, which is around 30% of total trips and rather biased. The trip origins are uniformly at random distributed to buildings in OpenStreetMap (without being able to weight differently for a small terraced house vs high-rise flats), and destinations are uniformly at random to buildings with some kind of commercial amenity.

(For non-UK maps, I think the only other travel demand model built in is for Seattle, and it comes from https://www.psrc.org/activity-based-travel-model-soundcast, but also quite old data at this point.)

This impact prediction tool is just taking all of the driving trips from the above model and calculating the shortest driving route, assuming free-flow traffic everywhere, and without any guesses about things like delays at traffic signals. Then after you make changes to a neighbourhood, it calculates the same thing again but using those changes, and shows you the difference. It's very rough, but hopefully gives a general sense of how traffic might divert to nearby areas in the short-term. You can see if the optimal routes tend to be on main roads or if another neighbourhood might also need some interventions.